1 

Trade Restrictiveness Indices in Presence of Externalities:

An Application to Non-Tariff Measures

John Christopher Beghina Anne-Celia Disdierb Stéphan Marettec

a: Iowa State University, 383 Heady Hall, Ames, IA 50011, United States. Email:

beghin@iastate.edu

b: Corresponding author: Paris School of Economics, INRA, 48 boulevard Jourdan, 75014 Paris,

France. Tel: +33 1 43 13 63 73. Email: Anne-Celia.Disdier@ens.fr

c: INRA, UMR Économie publique, Avenue Lucien Brétignières, 78850 Thiverval-Grignon,

France. Email: stephan.marette@agroparistech.fr

  2 

Trade Restrictiveness Indices in Presence of Externalities:

An Application to Non-Tariff Measures

Abstract:

We extend the trade restrictiveness index approach to the case of market imperfections and

domestic regulations addressing them. We focus on standard-like non-tariff measures (NTMs)

affecting cost of production and potentially enhancing demand by increasing product quality or

reducing negative externalities. We apply the framework to the database of Kee et al. (2009) and

derive ad valorem equivalents (AVEs) for NTMs. Half of the product lines affected by NTMs

exhibit negative AVEs, indicating a net trade-facilitating effect of NTMs. Accounting for these

effects significantly reduces previous measures of countries’ trade policy restrictiveness obtained

while constraining NTMs to be trade reducing.

Keywords: Non-tariff measures, externalities, ad valorem equivalents, trade restrictiveness

indices

JEL code: F13, F18, Q56

  3 

1. Introduction

Standard-like non-tariff measures (NTMs) are playing an increasing role in international trade.

Some of them have protectionist purposes, especially in a context of decreasing tariff barriers.

However, some others are adopted by policymakers to address market imperfections

(externalities, information asymmetries). In such cases, NTMs may be trade facilitating and

welfare enhancing. The literature measuring the restrictiveness of the trade policy, through the

computation of various indices, has failed to consider these effects. Our paper fills this gap.

With global sourcing, it becomes challenging to guarantee products’ safety and quality

and to mitigate negative externalities. Standards and regulations affecting quality help overcome

asymmetric information issues. Occasional recalls by toy, pharmaceutical and food companies

illustrate the importance of various safety concerns, such as led paints in children toys (Lipton

and Barboza, 2007). Consumers may also care about global commons and avoid purchasing

products obtained using unsustainable environmental practices. To preserve their reputation,

large firms (e.g. Home Depot, IKEA, etc.) have shown strong support for forest certification

(McDermott and Cashore, 2009). Similarly, consumer welfare is improved by quality

requirements limiting residues of dangerous pesticides and antibiotics in food products (Disdier

and Marette, 2010).

In this context, regulatory interventions have strong economic and political support,

despite risks of inefficiency and distortions. The effects of these regulatory instruments are

indeed complex not only because instruments are imperfect but also because they impact costs of

heterogeneous foreign and domestic producers. Meeting the NTMs is costly for both domestic

and foreign suppliers and often more so for the latter. While a regulation may thwart a market

failure and facilitate trade between countries, it may also reduce market access for foreign

  4 

producers who cannot easily comply with this regulation. To illustrate, between October 2006

and 2007, the U.S. Consumer Product Safety Commission (CPSC) announced 473 products

recalls of which 389 cases involved imported products (CPSC, 2008). This last effect may

outweigh the “legitimate action” to mitigate a market failure. Both trade and welfare impacts of

regulation are ambiguous and in general hard to evaluate. A rigorous empirical measure of these

impacts therefore requires a consistent framework, as proposed here.

We consider a small open economy, distorted, first, by arbitrary tariffs and other

domestic price policy distortions, and second by market imperfections and existing NTMs

allegedly addressing them. We pay particular attention to NTMs and their protective effects

against import competing products, as well as their potential demand enhancing effects when

NTMs reduce information asymmetries and trade cost. We then extend the trade restrictiveness

index (TRI) approach of Anderson and Neary (2005) to this more general and realistic case

encompassing market failures and the existing domestic regulations addressing them.

The TRI approach of Anderson and Neary (1992, 1994, 1996, 2003, and 2005) provides a

welfare-based consistent aggregation of various trade distortions into a scalar uniform surtax

factor, equivalent to these distortions in terms of their welfare effects. The TRI approach is a

concept applying to a whole economy because it relies on the balance of trade approach.

Nevertheless, it has been applied successfully to partial equilibrium and multi-market situations.

Feenstra (1995) has proposed some simplifying assumptions greatly fostering the applicability of

the approach by reducing the number of price responses to estimate or calibrate in the

implementation. The TRI and its extensions such as the Mercantilist TRI (MTRI) of Anderson

and Neary (2003) have been used to derive the tariff equivalent of arbitrary tariff structures

(Anderson and Neary, 1994), tariffs and quotas (Anderson and Neary, 1992 and 2005), tariffs

  5 

and domestic production subsidies (Anderson et al., 1995; Anderson and Neary, 2005; Beghin et

al., 2003), and tariffs and AVEs of other NTMs (Hoekman and Nicita, 2011; Kee et al., 2009;

Lloyd and MacLaren, 2008; and Bratt, 2012), among others. As shown in these applications, the

TRI approach provides a consistent aggregation of distortionary effects of various policy

instruments into a single “total” AVE within a given sector. The latter property explains the

recent success and popularity of the approach in empirical investigations of NTMs in presence of

tariffs and other price policies at the sector level.

The novelty of the present paper is to allow for market imperfections and trade

facilitating effects of NTMs in the TRI framework. Despite its inherent ability to capture second-

best situations, the determination of the TRI under market failure has been overlooked in the

trade literature. The only related effort in this direction is from Chau et al. (2007) who develop a

quantity-based distance function, a trade restrictiveness quantity index, in presence of

environmental externalities but abstracting from existing policy interventions. Outside of the TRI

literature, recent empirical investigations note that NTM regimes can facilitate trade (see Cadot

and Gourdon, 2013, for a review). Reputation and certification processes increase trust in

exchange (Blind et al., 2013); quality standards help reputation and reputation loss can be

detrimental to trade (Jouanjean, 2012); and transparency provisions in trade agreements can

facilitate regulated trade flows (Lejárraga et al., 2013).

We fill this gap in the TRI-related trade literature: we consider the TRI of arbitrary

tariffs, domestic production subsidies, and NTMs in presence of possible external effects.1 This

                                                            1 Several investigations using the standard gravity equation approach find some trade facilitating effects of NTMs

but without a rationalization based on some demand increasing effect or market imperfection presumably mitigated

by the NTMs being analyzed (see Li and Beghin, 2012).

  6 

undertaking is a substantive step forward for two reasons. First, trade policy reforms often occur

in the context of market imperfections such as asymmetric information or negative externalities

imposed on some agents. Accounting for these imperfections is relevant and has been the central

pillar of the trade and environment literature using the dual approach to trade (Copeland, 1994;

and Beghin et al., 1997). Surprisingly, this case has eluded the TRI literature. Second, numerous

NTMs have been emerging in the last 15 years for several reasons, including potential

protectionism, but also to address consumer and retailer concerns for health and the environment

and associated external effects. A priori, excluding potential market imperfections when

analyzing NTM policy reforms biases results and could lead to erroneous policy

recommendations. Not surprisingly, sectoral AVEs and TRI estimates are likely to exhibit

upward bias when they are econometrically constrained to treat all policies as trade-reducing. We

depart from this restrictive premise and start from an agnostic prior on the impact of NTM

policies on trade and welfare.

We then apply the proposed framework to the NTM global database of Kee et al. (2009)

consisting of a large cross section of products (at the 6-digit level of the Harmonized System –

HS – classification) and importing countries. We derive ad valorem equivalents (AVEs) for

NTMs and other policy distortions (tariffs and domestic production subsidies). 20% of HS 6-

digit lines are affected by NTMs and nearly half of these (10% of the lines) exhibit negative

AVEs of NTMs, indicating a net trade-facilitating effect of NTMs in those sectors. These AVEs

are then used to evaluate the restrictiveness of the trade policy defined by countries. TRIs

computed with these AVEs reflect the frequent trade facilitating effect of NTMs. Accounting for

these trade-facilitating effects significantly reduces previous measures of trade policy

restrictiveness for most countries obtained while forcing NTMs to be trade reducing. These

  7 

trade-facilitating effects cast doubt on the predominant presumption that NTMs are exclusively

protectionist and cannot possibly boost trade, let alone welfare.

Our paper proceeds as follows. We present the framework in Section 2. We then describe

the data and detail the econometric approach in Section 3. Section 4 presents the estimation

results of AVEs and TRIs. We conclude in Section 5.

2. The TRI framework with market imperfection

We follow the standard TRI approach with the balance of trade function derived from the dual

approach to trade for a small open distorted economy. We build on the usual framework with a

negative externality affecting the representative consumer as in Copeland (1994). The externality

is assumed exogenous to the consumer but influenced by the policymaker via some NTM

regulations such as standard-like regulations. These regulations may not be set optimally and

may be set at a protectionist level as in Fisher and Serra (2000).

2.1. Market demand and supply, and balance of trade function

The utility of the representative consumer is u(x,H(NTM)) with non negative market

goods x and negative externality H influenced by a vector of NTM policies, NTM, and with the

usual definitions and properties:2

.0/ with )(

;0/ and 0/

NTMHNTMHH

Huuxuu Hx

All domestic consumer prices p are inclusive of the exogenous world price wp, a tariff τ,

                                                            2 We could complicate the model by assuming that imports m influence the health externality or H(m(NTM), NTM).

This would make health depends on all the arguments influencing imports and generate clutter with multiple

feedback effects of all policies through health. The effect of NTM alone on health generates the possibility of trade

enhancements which is what we are after.

  8 

and the unit cost equivalent of the domestic NTM on foreign suppliers to sell in the domestic

market, or p = wp + τ + t(NTM).3

Given domestic prices p, the associated expenditure function is:

);|'(),,( HHuuxpMinHupex

,

with the usual derivative properties:

.0/ and ,0))(,,(/ HeeNTMHupxpee Hp

Expenditure function e exhibits all the usual homogeneity and curvature properties in

prices, implying p’epp=0, eH=p’epH, eu=p’epu ; epNTM = epH HNTM , and f’eppf ≤ 0 for any arbitrary

vector f of similar dimension as p. The marginal damage eH of the negative externality is positive

for any given utility level. To keep utility constant, expenditure has to increase when the

negative externality increases. Partial derivative eu is the inverse of the marginal utility of

income assumed positive. We eventually simplify preferences to follow Feenstra (1995) in the

empirical investigation section.

The impact of the NTM policy encompasses several possible cases. The demand

enhancing case is epNTM = epH HNTM < 0. Protectionism of the NTM is implied by HNTM = 0

because the policy does not address an externality or is not based on science. Another special

case could be that the NTM policy affects H (Hntm<0) but that H(NTM) does not affect a

particular demand (particular good n) directly, or epnH = 0. In this case, the policy is not

protectionist per se but addressing the market imperfection has no bearing on that particular

demand for good n. These last two cases show the difficulty to gauge revealed protectionism.4

                                                            3 Domestic and foreign firms have heterogeneous cost of meeting the NTM standard as explained later in the

production component of the model and we assume that domestic firms are more efficient at meeting these NTMs.

4 Demand not being enhanced by the NTM policy is not sufficient although suspicion of protectionism may arise.

  9 

For integrability of the Hicksian demands into the expenditure function, at least one of

the demands represented by x has to be influenced by the external effect H. To illustrate, H could

be the negative health effect of consuming products that are hazardous if minimum quality

standards are not imposed on their production. The standard reduces the occurrence of sickness

which may affect the demand for these products, and possibly other demands via better health

(reduced medical expenditure, more active leisure activities) or none other at all (all other

demands independent of health status). Similar examples can be constructed with environmental

external effects such as global commons or consumer packaging waste in retail consumption.

On the production side, domestic supply decisions in competitive industries are derived

from the gdp function:

( max(, ) ' ( , ) 0) ,p p

ygdp p p y y zz g

with y denoting the net output vector, z the vector of fixed national endowments, and pp the

vector of producer prices. Producer prices include production subsidies, s, such as farm

subsidies, not seen by consumers, ( )pp wp t NTM s . World prices can be normalized to 1

so the distortions s, t, and τ are viewed indifferently as either ad valorem or specific policy

distortions. For simplicity we assume that domestic firms already meet the standards implied by

NTM but that foreign firms may not. A more complicate framework affecting both domestic and

foreign firms could be included but the essence here is that t(NTM) captures the asymmetric

protective effect of NTM at the border on foreign industries.5 The gdp function has the usual

envelope and homogeneity properties:

                                                            5 NTM would then enter the GDP function and the derivative pNTM NTMgdp y would represent the leftward shift

of domestic supplies caused by the NTM policies. The unit cost equivalent of yNTM

would be assumed to be smaller

than t(NTM) to indicate a net protective effect of NTM on domestic suppliers as in Fisher and Serra (2000).

 10 

pgdp = 0; and ' 0 for a/ ; ' ; ny/ ' .'p p p p pp p ppp f ggdp p y p gdp gdp p y p p g p dp fd f

For convenience we also define compensated excess demand functions m, with

( , , ( ), , ) ( , , ( ) ( , )p pm p p H NTM u z x p u H NTM y p z , with partial derivatives indicated by the

appropriate subscript as for functions e and gdp.

Now we have all the elements to develop the balance of trade function B:

( , )

( ) '( ( , , ) )) ' ).

p

p p p

B p, p wp, NTM,z,H, u

e(p, u, H NTM )- gdp(p , z) - τ x p u H - y(p , z s y(p , z

(1)

Variable B indicates the amount of foreign exchange necessary to sustain utility u given NTM,

wp, z, s, and τ. Homogeneity in prices and envelope properties of e and gdp lead to a simpler

formulation of (1) seemingly omitting tariff revenues and production subsidy costs.

( , , , , ( ), ) (1 ( )) '( ( , , ( )) - ( , ).p pB p p wp z H NTM u t NTM x p u H NTM y p z (1’)

2.2. Trade restrictiveness indices with externality

The TRI problem in our case is to find a scalar T equivalent to standard-like policies, tariffs, and

production subsidies to apply as a tariff surcharge on world prices such that:

0

0 0 0 0 0 0 0 0

( (1 ), (1 ), , , (0), )

( ( ), ( ) , , , ( ), ) .

B wp T wp T wp z H u

B wp t NTM wp t NTM s wp z H NTM u B

(2)

The tariff surcharge accounts for several components: tariffs τ, domestic production

subsidies s, the demand shift via H(NTM), and the protective effect from raising foreign cost to

satisfy technical measure NTM, that is, t(NTM).

Next, while holding u constant, we differentiate equation (2) with respect to T, τ, s, and

NTM to derive the relative change in T rather than T as it is customarily done in the TRI

literature. This step yields:

 11 

' ' ' ' '( ) ( )( / ) ,p p pp p H NTMp p pB wp B wp dT B B d t NTM dNTM B ds B H dNTM

(3)

with subscripts denoting the variable involved in the partial derivative of B. Solving for dT

yields:

' ' ' ' ' ' '(1/ ( ))[( ) (( ) / + ) ],p p p pp p p H NTMp p p pdT B wp B wp B B d B ds B B t NTM B H dNTM

(4)

with partial derivatives Bi:

' '

' '

'

;

( ) ;

( ( )) 0.

p

p pp

ppp

H pH

B e

B s gdp

B wp t NTM e

Equation (4) shows that the TRI has three policy components corresponding to the tariff,

subsidy, and NTM policies. The NTM component is the sum of a demand effect via reduced

externality H, and a NTM protectionist effect relative to foreign goods (through a tariff

equivalent t increasing in NTM). While the sign of this protectionist effect on imports is clear, the

combined effect of NTM on m via the externality H and the protectionist effect t(NTM) is

ambiguous as their relative magnitude is unknown analytically. For example, a pure protectionist

NTM policy imposing useless labeling requirements would raise t(NTM) and have no effect on

consumers’ perception and would lead to a welfare loss and trade contraction. Conversely,

standards requiring safe goods including imported ones are likely to lead to a net demand-

enhancing effect lowering transaction costs for consumers. The latter NTM policy would be

trade and welfare enhancing. The econometric investigation will sort the NTM regimes into trade

reducing and trade facilitating since we do not impose any “protectionist” NTM prior.

Next, to further elucidate these effects and undertake our empirical investigation, we

assume a simplified structure for the Hessian matrix of cross-price responses (epp - gdppp) as in

Feenstra (1995), and others. The Hessians epp and gdppp are each assumed to be diagonal and

 12 

constant, which leads to ' '0 and 0 if and are non negativepppB B s .6 From these conditions

we derive an implementable framework to approximate the sector total AVE corresponding to all

policy types τ, s and NTM as well as the implied TRI and the MTRI. In general, if the Hessian

matrices of price responses of imports (or demand and supply responses) are not constrained to

be diagonal, off-diagonal elements can be positive or negative and it is impossible to a priori

sign elements of and pp pB B and therefore the change in the TRI, dT. The computation of T is

obviously cumbersome in the presence of cross-price effects and non-constant slopes.

We recover TRI T from dT as in Feenstra (1995) and Kee et al. (2009), which is

equivalent to the initial tariffs, subsidies, and NTMs relative to a world with all policies set to 0

by integrating both sides of (4) with respect to T going from zero to T and policies going from

(0,0,0) to (τ, s, NTM). The latter approach works only if dT is non-negative. This step yields:

(1/ '( )) ( ) ,p ppp pp p p NTMpT wp gdp e wp B B B B TMs N (5)

with ' '( ) / +pNTM p H NTMpB B B t NTM B H whose sign is undetermined. The original formula in

Feenstra (1995) contains the first positive element from tariffs abstracting from s and NTM.

Here, two additional components originate from production subsidies (positive contribution to

the TRI), as long as subsidies are positive, and from NTM policies (ambiguous sign). The

formula in Kee et al. (2009) has the protectionist effects of tariffs and subsidies and a

protectionist effect of NTMs. No externality or demand enhancement appears in their equation.

This additional effect included in our equation (5) can potentially facilitate trade and complicates

the simple narrative of obstructive NTM policies and their tax equivalent. Equation (5) is in

                                                            6 This simplification reduces price effects to the own-price effect, and homogeneity holds implicitly by defining

prices relative to a numéraire good.

 13 

essence the square root of a weighted sum of deadweight losses from tariff, production subsidies,

and the welfare effects of NTMs. If the latter is a pure protectionist policy, then BHHNTM is zero

(no demand shift) and the dead weight loss from the tariff equivalent t(NTM) is added to the sum

of deadweight losses. If the NTM policy facilitates trade, then the latter maps into a welfare gain.

Removing the NTM decreases the TRI as welfare falls with its removal. If the latter effect

dominates the distortionary effect of tariffs and subsidies, then dT is negative and T cannot be

recovered using (5). Instead, dT is the form of choice as in the early TRI investigations (e.g.,

Anderson et al., 1995).

These effects are illustrated in partial equilibrium in Figure 1. Figure 1 shows the two

effects of the NTM policies, that is, the demand enhancement shift (from x to x’ with greater

utility achieved with reduced health hazard), and the increase in border price (wp+t(NTM)+τ)

reflecting the international cost of meeting the country’s standard and the tariff, and their total

effects on imports m. In previous investigations only the border price effect of NTM, t(NTM),

was considered and the trade (and welfare) impact of NTM on imports was detrimental by

assumption.

Insert Figure 1 here

Along with the TRI, we consider the MTRI, which holds aggregate imports (wp’m)

constant. The MTRI yields the tariff equivalent to all distortions holding aggregate trade

unchanged but allowing for welfare variation. The MTRI is derived in Anderson and Neary

(2003) and Kee et al. (2009) who call it the overall TRI (OTRI). The derivation of the MTRI

follows the spirit of the derivation of the TRI and we only present its final formula in equation

(12). We refer readers to Anderson and Neary (2003) for details.

An important consequence from the potential presence of trade-enhancement effects and

 14 

negative AVEs from NTMs is that our TRI and MTRI estimates will be equal or smaller than the

TRI and MTRI where all policies are constrained to be trade reducing. We discuss this important

point in the empirical section.

2.3. The import equation to estimate

Next, we derive the import equation to estimate and the AVEs of all policy instruments. Totally

differentiation of m (holding u constant) for changes in exogenous variables leads to a change in

imports of good n in any country equal to:

( / ) ( / ) [( / )( / )

( / )( / )] ( / ) .

n n n n n n n n n n n

n n n n n n

dm m dp d y p ds m dp t NTM

x H H NTM dNTM y z dz

(6)

Equation (6) and m provide a way to estimate the response of imports to tariffs, subsidies,

and NTM policies, and other variables as in Feenstra (1995). We then derive the estimate of the

AVE to the net effect of NTM policies on good n. Unfortunately we cannot separately identify

the individual effects of NTM on m in (6), but we can estimate their net effect. Following a

common practice we move the tariff effect on the left hand side of (6) and the general

specification for the import demand of good n in country c (as indicated by superscript n,c) is:

, , , , ,, , ,ln ln(1 ) .n c n c z n c S n c NTM n c

n c n k k n c n ck

m z s NTM (7)

Elasticity n,c is the own-price response of import of good n in country c. ,NTMn c is the sum of two

AVE components (the tariff equivalent of NTM on world prices, and the ambiguous import

subsidy/tax effect of NTM via decreased externality). Note that the latter AVE component is

bound to the left to -100% as prices are non-negative. This non-negative constraint provides a

lower bound of -100% on cn

NTM, if we further assume that there is no trade impediment effect of

the NTM policy (t(NTM)=0) at the border. This is a limit case to establish the lowest non-

 15 

negative prices faced by agents in the economy.

Equation (7) once estimated provides the basis for the total AVE of NTM policies on

good n, NTMtotalAVE , which is:

, ,/ , with 1 .total

NTM NTM NTMn c n c totalAVE AVE (8)

An AVE is developed similarly for production subsidies, based on the fact that

cncnS

cnSAVE ,,, /)1( , with ( /

/= x pm p ). Unfortunately, parameter γ is not readily known as

we only have estimates of import demand price elasticities and not the underlying output and

demand price responses. Hence, we estimate a lower bound to the production subsidy AVE by

abstracting from fraction (1-γ). Alternatively, the production subsidy AVE estimate could be

seen as a market price support subsidy, affecting both consumer and producer prices. This

assumption is common although not fully accurate.

Next, we specify ,NTMn c as a transformation of an exponential such that it satisfies a lower

bound on the total AVE of the NTM effects as before and in addition allowing for fixed effects

per commodity and interaction terms with country-specific exogenous shifters (endowments) z.

For a continuous NTM variable, this leads to ,, ( )expNTM NTM NTM n c

n c n nk kk

a z , with parameter

a constrained such that the AVE of NTM is lower bounded at -1 or -100%. The corresponding

value is a=εn,c. If NTM is approximated by a dichotomous variable, then the various partial

derivatives of m, and t with respect to NTM do not exist and are replaced by the first difference

of m for NTM equal to one and zero. This leads to an alternative formula of the total NTM AVE

( dumNTM

totalAVE ) following Halvorsen and Palmquist (1980):

, ,[exp( ) 1] / , with 1 .dum dumNTM NTMNTMtotal n c n c totalAVE AVE

(9)

The lower bound condition in (9) is slightly more cumbersome with a dichotomous NTM.

 16 

The intuition is that ,exp( ) 1NTMn c

cannot be too large of a positive number to keep producer and

consumer prices non-negative (or that , ,exp( ) 1NTMn c n c or , ,ln(1 )NTM

n c n c ). Using the

same specification as for the continuous variable case of ,NTMn c , we specify the lower bound

constraint for the dichotomous case using parameter a in ,, ( )expNTM NTM NTM n c

n c n nk kk

a z with

,ln(1 )n ca . For small values of ,n c , the dichotomous and continuous values of a are

approximately equal.

A parallel formulation is used for k

cn

knk

S

n

S

cn

Sz )exp(

,

, . As production subsidy s

is positive, presumably its AVE would not lead to negative producer price issues.

The total AVE of all distortions, that is, tariffs, NTMs, and subsidies for good n in

country c is then (assuming the normalization wp=1):

.,,,, cn

s

cn

NTM

cncn AVEAVETOT (10)

The TRI in equation (5) translates into:

1/22,( / )

( / )

nc nc n cn

cnc nc

n

m p TOTT

m p

. (11)

Again, if (4) gives a negative dT, then (11) cannot be used and the change in TRI, dT, is

kept to express the change in the index equivalent to the welfare impact of the policy

interventions. Recall that dT is expressed as a sum of consumer welfare changes, and that T is the

square root of a positive sum of deadweight losses.

As noted above, we use the same data and AVE estimates to compute the MTRI, merccT :

 17 

,( / )

( / )

nc nc n cmerc n

cnc nc

n

m p TOTT

m p

. (12)

3. Data and econometric specification

We use the UNCTAD7-Comtrade database of Kee et al. (2009)8 as well as their import demand

estimates (Kee et al., 2008) to estimate the import demand equation (7), recover AVEs

(equations (9) and (10)) at the 6-digit level of the Harmonized System (HS), and compute the

MTRI and TRI, (and dTRI) equivalents to the three types of distortions (tariffs, NTMs and

subsidies) as in equations (11) and (12) (or (4) for negative dTRI) for each country.

3.1. Data

Trade data come from the Comtrade database. We use the average of imports at the HS 6-digit

line by importing country between 2001 and 2003. Imports demand elasticities are extracted

from Kee et al. (2008). Tariff data are taken out from the UNCTAD and the World Trade

Organization (WTO). Tariffs are for the most recent year for which data are available between

2000 and 2004. For specific tariffs, ad valorem equivalents are used. Data on NTMs are from the

UNCTAD TRAINS (Trade Analysis and Information System) database and the following NTMs

are selected: price control measures, quantity restrictions, monopolistic measures, and technical

regulations. A dummy is set to one if the importing country imposes at least one NTM on a given

HS6 product. Regarding production subsidies, the dataset of Kee et al. (2009) covers agricultural

                                                            7 United Nations Conference for Trade and Development.

8 As recently pointed by Breaux et al. (2013), the new NTM data collection effort under the interagency MAST

project seems to be problematic and less promising than one could have hoped. The older TRAINS database appears

more reliable than the new MAST dataset.

 18 

domestic support. The source is the WTO domestic agricultural support notifications. This

continuous variable is in dollars and its log form is used in the estimations.

Countries’ characteristics are measured by the economic size (gross domestic product –

GDP), and relative factor endowments (agricultural land over GDP, capital over GDP, and labor

over GDP). Data are extracted from the World Development Indicators of the World Bank. Two

geographical variables are also introduced: a dummy for islands and a measure of remoteness

(average distance to world markets defined as the import-weighted distance to each trading

partner). Our sample includes 93 importing countries and 4,941 products (HS6 lines).

3.2. Econometric specification

We run estimations HS 6-digit line by HS 6-digit line. To control for the potential endogeneity of

NTMs and production subsidies, we instrument them using exports, GDP-weighted average of

the NTM dummy variable at the HS 6-digit of the 5 closest neighbors (in terms of geographic

distance) and the GDP-weighted average of the agricultural domestic support at the HS 6-digit of

the 5 closest neighboring economies (Kee et al., 2009). The instrumented estimation is

performed in two stages. We first estimate a probit where the dependent variable is the presence

or the absence of a NTM and the explanatory variables are the instruments. The mills ratio

derived from this first stage is then included in the second stage equation. If one (or more)

country provides production subsidies, instruments for this variable (exports, GDP-weighted

average of the agricultural domestic support of the 5 closest neighbors) are also included in the

second stage equation.

The quantity impact of NTMs and production subsidies is then transformed into price-

equivalents (AVEs) using the provided import demand elasticities. AVEs are calculated for each

 19 

importing country and HS6 line. We impose a positive cap AVEs at 50 for a few extreme values.

To ease result interpretation, we compute the mean over all importing countries at the HS6 and

HS2 levels. Following our estimation, 10% of AVEs for NTMs at the HS 6-digit level are

negative, i.e., highlighting trade-facilitating NTMs. Without constraint on the sign of the AVEs,

our procedure allows us to keep these negative values in our sample. AVEs of NTMs, tariffs and

production subsidies are then aggregated at the country level to derive the trade restrictiveness

indices corresponding to all three types of policy interventions.

Finally, we use bootstrapping to compute the standard deviations of the AVEs. The main

advantage of this procedure is to account for sampling and estimation errors of the AVEs. We

draw (with repetition) 200 random samples from our dataset and perform the AVEs estimation

for each of these samples. Estimations are run HS6 line by HS6 line. We then compute the

bootstrap standard errors as the standard deviations of these 200 AVEs.

4. Results

We first present the results on AVEs of NTMs in the presence of externalities. We also provide

comparisons with the AVEs obtained when the latter are constrained to be trade reducing.

4.1. AVEs of NTMs

We focus the discussion on the results obtained for the first 20 HS sections.9 Qualitative

conclusions are unchanged if the discussion of results is performed at the HS 2-digit level (with

96 sectors, see Table A.1 of the Online Appendix attached for review). Table 1 first reports the

simple frequency ratio of NTMs for each HS section, i.e., the share of HS6 lines within each HS

section for which at least one importing country of our sample imposes at least one NTM. The

                                                             9 Section XXI (objects of art and antiques) has very few HS6 lines with NTMs and is not reported.

 20 

frequency ratio of NTMs should be interpreted as follows: for section I “live animals, animal

products”, the value 0.458 means that 45.8% of HS6 lines included in HS section I are affected

by at least one NTM in at least one importing country.

Results suggest that agricultural and food products (sections I through IV) are more

affected by NTMs than manufactured products. The frequency ratio is indeed larger for these

products. These industries have high numbers of countries’ notifications of sanitary and

phytosanitary measures to the WTO. According to the results presented in the Online Appendix

(Table A.1), for some HS 2-digit sectors, such as live animals, meat, dairy products, edible fruit

and nuts, more than half of the HS6 lines are subject to at least one NTM in one importing

country. By contrast, for a number of manufactured products, the share of HS6 lines impacted by

a NTM is lower to much lower. A strong exception is “pharmaceutical products (HS30)”

(frequency ratio of 52.7%). Many chemical and allied industries (section VI) have frequencies

between 15 and 30%. Interestingly, textiles and apparel (section XI) and footwear and headgear

(section XII) for which the competition between Northern and Southern countries has been

historically contentious, are subject to many NTMs suggesting that some of them may be

protectionist measures.

The next column of Table 1 reports the average AVE of NTMs for each HS section

allowing for the presence of externalities. The mean is computed over all importing countries

and HS6 lines within each section. The mean AVE on the whole sample is equal to 0.035, but

strong differences can be observed across sections. First, the magnitude of the mean AVE varies

significantly across sectors and is much higher for agricultural products and footwear/headgear

than for other products. Second, almost all sections exhibit a positive average AVE, indicating

that NTMs have, on average, a net negative impact on trade flows. However, for three sections

 21 

(chemicals and allied industries, pearls and precious metals and stones, and arms and

ammunition10), the average AVE is negative, suggesting that NTMs are trade-facilitating either

by improving quality, reducing information asymmetries or by being anti-protectionist. Not

accounting for these positive trade effects will therefore bias the computation of AVEs, TRIs,

and MTRIs. In our sample, 20% of HS6 lines are affected by NTMs and half of them exhibit

negative AVEs of NTMs. These negative AVEs are spread over all HS sections (and HS2 sectors

as shown in Table A.1 of the Online Appendix). Column (3) of Table 1 underlines the upward

bias affecting the estimation of AVEs when NTM are constrained to be trade-reducing. As

expected, the average AVE for each HS section is systematically higher than the average AVE

obtained in column (2).

As highlighted with the frequency ratio, the share of HS6 lines subject to at least one

NTM greatly differs across section and could therefore bias the average AVE calculated using all

HS6 lines. To control for this bias, columns (4) and (5) of Table 1 report the average AVE

computed only on HS6 lines on which at least one NTM is applied. Column (4) allows for the

presence of market imperfections and trade-facilitating NTMs, while column (5) does not. As

expected, the average AVE computed only on HS6 lines subject to a NTM is always higher in

absolute value than the one based on all HS6 lines (with or without a NTM). However, the

ranking of sections is now slightly different. AVEs of NTMs are still high for several agricultural

products (especially for fats and oils, and live animals and animal products). However, the

magnitude of the mean AVE is also notable for some manufactured products (e.g. machinery,

electrical and video equipment). Furthermore, the difference between the AVEs computed using

                                                            10 The sector of arms and ammunition is least likely to observed commercial trade and standards like NTM policies.

Most of the negative AVES are for 9306 sub-sectors including cartridge for pellet guns and sports guns.

 22 

all HS6 lines and using only lines with a NTM cannot only be explained by the frequency of

NTMs. For example, the frequency ratio of NTMs is relatively similar for pulp of wood, paper

and printing (section X, ratio: 13.1%) and optical, photographic and medical instruments (section

XVIII, ratio: 13.2%). However, the difference between the average AVE based on HS6 lines

subject to a NTM and the one based on all HS6 lines is higher for optical, photographic and

medical instruments than for pulp of wood, paper and printing (0.489 vs. 0.423 in the constrained

estimation and 0.089 vs. 0.061 in the unconstrained one). This result is also observed at a more

disaggregated level (see Table A.1 in the Online Appendix). This divergence of AVEs can be

rationalized by the difference in the shares of trade reducing and facilitating NTMs across

sections as well as in the magnitudes of the AVEs of these NTMs.

Insert Table 1 here

Table 2 distinguishes between trade-reducing and trade facilitating NTM estimates using

results from the unconstrained estimation (allowing for external effects). Again results are

summarized by HS section. The first column of Table 2 provides the share of NTM-ridden

observations with positive AVEs (trade-reducing NTMs). This share varies across sections, from

18.3% (arms and ammunition) to 65.3% (fats and oils). For 15 out of 20 sections however, the

majority of NTMs are trade-reducing (with a share above 50%). In total, 51.8% of NTM-ridden

lines at the HS6 level are negatively affected by NTMs.

The last 2 columns of Table 2 show the mean AVE for trade-reducing NTMs and that of

trade-facilitating NTMs by HS section. We previously noticed that NTMs were more numerous

on agricultural products. According to the second column of Table 2, the AVEs of trade-reducing

NTMs on agricultural and food products are however not necessarily higher than the ones

 23 

obtained on manufactured products. For example, the average AVE for mineral products (1.279)

is slightly larger than the ones observed for vegetable products (1.047) or prepared foodstuffs,

beverages, spirits and tobacco (1.130). The average positive AVE for the whole sample is equal

to 1.111. In the last column of Table 2, AVEs of trade-facilitating NTMs are non positive, and

because of the non-negative price constraint, they are included in the interval [-1;0]. Interestingly

we observe that the magnitude of these AVEs is high in absolute value. The minimum in

absolute value per section is equal to -0.803 (prepared foodstuffs, beverages, spirits and

tobacco), and the maximum (-0.974) is reached for pearls, precious metals and stones. The mean

over all sections is -0.840. Table A.2 of the Online Appendix presents the results at the HS 2-

digit level. Previous conclusions are still valid and some heterogeneity is also observable across

HS2 sectors in the magnitude of the AVEs of trade reducing and facilitating NTMs.

Insert Table 2 here

Figures 2 and 3 provide further insights on the NTM AVES. Figure 2 shows the scattered

plot of AVEs at HS6 level, average over all countries and sorted by HS2 line (x-axis numbered

from 1 to 96 for 96 HS2 lines). The plot shows the non-negative price constraint (lower bound at

-1) and the density of negative (and positive11) AVEs for most HS2 lines, and in particular for

fish and crustaceans (line 3), inorganic and organic chemicals (lines 28 and 29), and iron and

steel and articles of iron and steel (lines 72 and 73), nuclear reactors, electrical machinery and

equipment (lines 84 and 85), and optical, photographic, measuring, precision and medical

instruments (line 90). The plot also shows the presence of large positive outliers for many HS6

lines. Figure 3 shows two central values (median and mean) of the HS6 AVE averages by HS2

                                                             11 The plot is truncated from above at AVE=3 for better clarity but misses less than 0.3% of the AVE estimates.

 24 

line. Most of the within-HS2 means are higher than the corresponding medians, which is

motivated by the constraint on non-negative prices and the presence of large positive outliers.

Many medians and some means are negative suggesting again the presence of a number of trade-

facilitating NTM regimes in sector like food, chemicals, and precious stones and metals. To

offset that, positive AVEs also abound suggesting trade-reducing effects in various sectors most

visibly in dairy products (line 4), various textiles and apparel, and footwear and headwear (lines

64 and 65). These sectors are known for their history of protectionism in many countries.

To sum up, our results suggest the presence of both trade reducing and facilitating

NTMs, with substantial trade effects. Next, these AVEs of NTMs are further used to calculate

the TRI and MTRI.

Insert Figures 2 and 3 here

4.2. Trade restrictiveness indices

Table 3 reports summary figures of the results for country-level MTRIs, TRIs and changes in

TRIs. Three calculations are performed based on (i) tariffs only, (ii) overall protection using

AVEs from the constrained estimation, and (iii) overall protection using unconstrained AVEs.

The latter two sets of measures are also summarized for all AVE estimates and for the subset of

significant AVE estimates based on the bootstrap standard errors. The summary statistics are

presented for all 93 countries, OECD countries, Least Developed Countries (LDCs), and then

BRIC (Brazil, Russia, India and China) countries.

The tariff only MTRI and TRI (1st and 6th columns in Table 3) represent the uniform

tariff that would provide the same level of imports (MTRI) and welfare (TRI) as the initial tariff

structure. OECD countries where in most cases except Japan and South Korea tariffs have been

 25 

significantly reduced, exhibit smaller tariff-MTRIs than the 93-country averages, LDCs’ and

especially the BRICs’ averages. According to detailed country results reported in Table A.3 of

the Online Appendix, India has the highest tariff-MTRI (0.257) among the 93 countries; South

Korea and Brunei have the highest tariff-TRI at or above 0.5. Hong-Kong and Singapore have

zero tariff indices as they do not impose border tariffs.

Columns (2) and (7) show the MTRI and TRI estimates including all distortions based on

the AVEs from the estimation constraining NTMs to be trade reducing. As expected, MTRIs and

TRIs exhibit larger values than in columns (1) and (6) than those obtained using AVEs from the

unconstrained estimation (see columns (4) and (9)). For example for the 93-country summary,

the median and mean values of the MTRIs are respectively 0.133 and 0.167 with constrained

estimates and only 0.019 and 0.011 with unconstrained estimates. Similarly, for the TRI the

median and mean values are 0.350 and 0.357 (constrained estimation) versus 0.220 and 0.255

(unconstrained estimation). In other words, for all countries included in our sample, the MTRIs

based on overall protection (tariffs, production subsidies, and NTMs) and allowing for negative

AVEs are equal or smaller than the MTRIs based on overall protection computed with the

constrained AVEs. This last result suggests that some NTM regimes have trade facilitating

effects for most countries. Finally, regardless of the estimation method, when comparing results

using all AVE estimates or only the significant ones based on the bootstrap standard errors, one

notes with the latter that ranges are reduced for all indices except one (MTRI under the

unconstrained estimation approach).

Countries’ grouping also highlights interesting patterns. The OECD group exhibit

negative MTRI values with a mean near zero. The LDC group also exhibit negative MTRI values

in its range. This result may seem surprising but is consistent with the integration of LDCs in

 26 

European trade following a sequence of structural adjustment policies that removed many

protectionist NTMs and expanded preferential trade agreements. The latter induced upgrades of

SPS regulations and improved food safety in countries like Sénégal and Côte d’Ivoire among

others (FAO, 2003; Colen et al., 2012; and Maertens et al., 2012). Intuitively, many countries

with low tariff-MTRIs exhibit negative total MTRIs because small tariffs do not counterbalance

negative NTM AVEs.

Lastly, using more disaggregated results by country (see Table A.3 of the Online

Appendix), we note that only 14 over 93 countries the MTRI values including overall protection

based on unconstrained estimates are higher than the values based on tariffs only. If we abstract

from production subsidies from the computation, the share is even smaller (only 9 countries over

93). 12 The analysis of the TRIs shows 45 countries with total TRIs smaller than the tariff-only

TRI based on unconstrained estimates. These results show that positing protectionism NTMs

strongly biases the evaluation of the restrictiveness of NTM trade policies.

As previously mentioned, if equation (4) provides a negative dT (cf. supra), then the TRI

level T cannot be computed using (5). The last columns of Tables 3 report the change in TRI, dT,

i.e., the change in the index equivalent to the welfare impact of the policy interventions.

Country-level results suggest that for 27 over 93 countries, the change in TRI is negative (Table

A.3 of the Online Appendix). Furthermore, for 45 over 93 countries, these values are smaller

than the ones obtained when tariffs only are included in the computation (column (7) of Table

A.3). These two last results highlight that some NTMs can have positive welfare effects. Not

surprisingly, Singapore, Hong-Kong, and many OECD countries exhibit negative dTRIs. This

                                                            12 For three countries, the AVEs of farm subsidies are larger under the unconstrained estimation than under the

constrained estimation. 

 27 

result is consistent with Disdier et al. (2008)’s results showing intra-OECD agri-food trade being

enhanced by NTM regimes. Singapore’s consumer valuation for risk-averting regulations and

orderly markets has been documented elsewhere (Tan, 1999). Several LDC countries also exhibit

negative dTRIs and these can be rationalized by opportunities created with the agri-food trade

integration and policy reforms as noted earlier.

Insert Table 3 here

5. Conclusion

We extend the TRI approach to a small distorted open economy to account for market

imperfections (externalities, asymmetric information) and NTM domestic regulations addressing

them. Up to date, the presence of externalities and potential anti-protectionist effects of NTMs

has been ignored in TRI application. Allowing for such occurrence, we derive the AVEs of

NTMs, as well as the TRIs and MTRIs equivalent to all policy interventions (tariffs, NTMs and

production subsidies). We show that in general the impact of NTMs on import demand is

ambiguous depending on the relative strength of the import-facilitating effects of NTMs via a

shift in import demand, and the protective effect of the same NTMs at the border. We then apply

the approach to the UNCTAD-Comtrade database built by Kee et al. (2009). In our sample, 20%

of HS6 lines are affected by NTMs and about half of these (10% of all HS6 lines) show negative

AVEs of NTMs. The MTRI and TRI results show the non trivial sizeable changes in estimated

aggregate trade and welfare effects of existing trade policies. Policy recommendations on the

impacts of NTMs will be biased by overstating their trade reducing and welfare decreasing

effects.

Although we show it is possible to rationalize and econometrically identify trade-

facilitating effects of NTMs mitigating external effects and other market imperfections or having

 28 

anti-protectionist effects on domestic suppliers, we do so using relatively simple NTM proxies. It

would be interesting to refine these results and use more detailed NTM measures and focus on a

subset of sectors for which we identify negative NTM AVEs. Nevertheless our results

corroborate the trade-facilitating effects found in the literature for some products and countries

(e.g. Disdier et al., 2008; Moenius, 2004). The value added of our analysis is to formalize the

possibility of anti-protectionist effects or external effects and their mitigation through regulations

affecting quality of products and identify their effects on trade restrictiveness. Our analysis also

extends the applicability of the TRI framework to more plausible market conditions and lets the

data reveal unconstrained patterns.

Acknowledgements

With the usual disclaimers, we thank Hiau Looi Kee, Alessandro Nicita and Marcelo Olarreaga

for providing their dataset, and for discussions, along with Rick Barichello, Antoine Bouët, Jean-

Christophe Bureau, Guillaume Gruère, Michael Ferrantino, Julien Gourdon, Dermot Hayes,

Daniel Sumner, Frank van Tongeren, and participants at the ETSG 2012 Leuven 14th Annual

Conference, KUL, Leuven, Belgium; the 2012 Paris Environmental and Energy Economics

Seminar series; the 2012 Iowa State University economics departmental seminar series; the 2012

International Agricultural Trade Research Consortium Annual Meetings in San Diego,

California; the 2013 Global Trade Analysis Conference in Shanghai, China; and the 2013

EAERE Conference in Toulouse, France.

 29 

References

Anderson, J.E., G.J. Bannister, and P. Neary (1995). “Domestic Distortions and International

Trade,” International Economic Review 36(1): 139-157.

Anderson, J. and P. Neary (1992). “Trade reforms with quotas, partial rent retention and tariffs,”

Econometrica 60(1): 57-76.

Anderson, J. and P. Neary (1994). “Measuring the restrictiveness of trade policy,” World Bank

Economic Review 8(2): 151-169.

Anderson, J. and P. Neary (1996). “A new approach to evaluating trade policy,” Review of

Economic Studies 63(1): 107-125.

Anderson, J. and P. Neary (2003). “The mercantilist index of trade policy,” International

Economic Review 44(2): 627-649.

Anderson, J. and P. Neary (2005). Measuring the Restrictiveness of Trade Policy, Boston: MIT

Press.

Beghin, J.C., J.C. Bureau, and S.J. Park (2003). “Food Security and Agricultural Protection in

South Korea,” American Journal of Agricultural Economics 85(3): 618-632.

Beghin, J.C., D. Roland-Holst, and D. van der Mensbrugghe (1997). “Trade and Environment

Linkages. Piecemeal Reform and Optimal Intervention,” Canadian Journal of Economics

30(2): 442-455.

Blind, K., A. Mangelsdorf, and J.S. Wilson (2013). “Mutual Recognition of Accreditation: Does

it Matter to Trade? Evidence from the Food, Beverage, and Tobacco Industry,” chapter 12 in

J.C. Beghin (ed.) Nontariff Measures with Market Imperfections: Trade and Welfare

Implications (Frontiers of Economics and Globalization, Volume 12), Emerald Group

Publishing Limited, pp. 291-310.

 30 

Bratt, M. (2012). “Estimating the Bilateral Impact of NTMs,” Manuscript, University of Geneva. Breaux, M., Y. Cabral, M.J. Ferrantino, and J.E. Signoret (2013), “Quality-Adjusted Handicraft

Estimates of NTM Price Gaps,” Manuscript, U.S. International Trade Commission.

Cadot, O. and J. Gourdon. (2013). “Assessing the price-raising effect of non-tariff measures,”

Manuscript, University of Lausanne.

Chau, N.H., R. Färe, and S. Grosskopf (2007). “Trade Restrictiveness and Pollution,” WP 2007-

15, Department of Applied Economics and Management, Cornell University, Ithaca, New

York.

Colen, L., M. Maertens, and J. Swinnen (2012). “Private Standards, Trade and Poverty:

GlobalGAP and Horticultural Employment in Senegal,” The World Economy 35(8): 1073-

1088.

Copeland, B.R. (1994). “International Trade and the Environment: Policy Reform in a Polluted

Small Open Economy,” Journal of Environmental Economics and Management 26(1): 44-65.

CPSC, U.S. Consumer Product Safety Commission (2008). “Import Safety Strategy,”

Manuscript, Washington D.C.

Disdier, A.-C., L. Fontagné, and M. Mimouni (2008). “The Impact of Regulations on

Agricultural Trade: Evidence from the SPS and TBT Agreements,” American Journal of

Agricultural Economics 90(2): 336-350.

Disdier, A-C. and S. Marette (2010). “The Combination of Gravity and Welfare Approaches for

Evaluating Non-Tariff Measures,” American Journal of Agricultural Economics 92(3): 713-

726.

FAO (2003). WTO Agreement on Agriculture: The Implementation Experience - Developing

Country Case Studies. Commodity Policy and Projections Service Commodities and Trade

 31 

Division FAO, Rome, 2003. 

Feenstra, R.C. (1995). “Estimating the effects of trade policy,” in G. Grossman and K. Rogoff

(eds.), Handbook of International Economics, vol. 3, Elsevier, Amsterdam.

Fisher, R. and P. Serra (2000). “Standards and protection,” Journal of International Economics

52(2): 377-400.

Halvorsen, R. and R. Palmquist (1980). “The Interpretation of Dummy Variables in

Semilogarithmic Equations,” American Economic Review 70(3): 474-475.

Hoekman, B. and A. Nicita (2011). “Trade Policy, Trade Costs, and Developing Country Trade,”

World Development 39(12): 2069-2079.

Jouanjean, M.A. (2012). “Standards, reputation, and trade: evidence from US horticultural

import refusals,” World Trade Review 11(03): 438-461.

Kee, H.L., A. Nicita, and M. Olarreaga (2008). “Import Demand Elasticities and Trade

Distortions,” Review of Economics and Statistics 90(4): 666-682.

Kee, H.L., A. Nicita, and M. Olarreaga (2009). “Estimating Trade Restrictiveness Indices,” The

Economic Journal 119: 172-199.  

Lejárraga, I., B. Shepherd, and F. van Tongeren. (2013). “Transparency in Nontariff Measures:

Effects on Agricultural Trade,” chapter 4 in J.C. Beghin (ed.) Nontariff Measures with

Market Imperfections: Trade and Welfare Implications (Frontiers of Economics and

Globalization, Volume 12), Emerald Group Publishing Limited, pp. 99-125.

Li, Y. and J.C. Beghin (2012). “A Meta-Analysis of Estimates of the Impact of Technical

Barriers to Trade,” Journal of Policy Modeling 34(3): 497-511.

Lipton, E. and D. Barboza (2007). “As More Toys Are Recalled, Trail Ends in China,” The New

York Times, Web Edition, June 20th.

 32 

Lloyd, P. and D. MacLaren (2008). “An Estimated Trade Restrictiveness Index of the Level of

Protection in Australian Manufacturing,” Australian Economic Review 41(3): 250-259.

Maertens, M., B. Minten, and J. Swinnen (2012). “Modern Food Supply Chains and

Development: Evidence from Horticulture Export Sectors in Sub‐Saharan Africa,”

Development Policy Review 30(4): 473-497.

McDermott, C. and B. Cashore (2009). “Forestry Driver, Mapping Project - Global and US

Trade Report,” Yale University, School of Forestry & Environmental Studies, GISF Research

Report 012.

Moenius, J. (2004), “Information versus Product Adaptation: The Role of Standards in Trade,”

Northwestern University, Kellogg School of Management, International Business and

Markets Research Center Working Paper.

Tan, S. J. (1999). “Strategies for reducing consumers’ risk aversion in Internet shopping,”

Journal of Consumer Marketing 16(2): 163-180.

 33

Table 1: Frequency ratios and AVEs of NTMs, by HS section

HS section codes

HS section names Simple

frequency ratio of NTMs

AVE of NTMs all HS6 lines (mean)

AVE of NTMs if NTM=1 (mean)

Unconstrained

estimationa Constrained estimationb

Unconstrained estimationa

Constrained estimationb

I Live animals, animal products 0.458 0.185 0.361 0.405 0.788 II Vegetable products 0.420 0.100 0.265 0.239 0.632 III Fats and oils 0.370 0.212 0.348 0.573 0.942 IV Prepared foodstuffs,

beverages, spirits, tobacco 0.422 0.126 0.302 0.299 0.715

V Minerals 0.096 0.026 0.069 0.266 0.722 VI Chemicals, allied industries 0.196 -0.030 0.091 -0.151 0.467 VII Plastics, rubber 0.160 0.036 0.105 0.223 0.655 VIII Hides, leather, furskins 0.123 0.019 0.072 0.150 0.584 IX Wood and wood articles 0.160 0.033 0.089 0.205 0.552 X Pulp of wood, paper, printing 0.131 0.009 0.063 0.070 0.486 XI Textiles, apparel 0.276 0.032 0.161 0.117 0.581 XII Footwear, headgear 0.239 0.095 0.169 0.399 0.708 XIII Stone, cement, ceramic

articles, glass 0.109 0.031 0.074 0.287 0.678

XIV Pearls, precious metals and stones

0.015 -0.005 0.004 -0.364 0.273

XV Base metals and articles 0.120 0.016 0.067 0.129 0.557 XVI Machinery, electrical and

video equipment 0.174 0.059 0.121 0.339 0.695

XVII Vehicles, aircraft, vessels 0.198 0.014 0.113 0.073 0.571 XVIII Optical, photo., medical instr. 0.132 0.014 0.074 0.103 0.563 XIX Arms, ammunition 0.306 -0.191 0.057 -0.625 0.186 XX Miscellaneous (furniture, toys,

others) 0.144 0.057 0.111 0.398 0.769

All sections 0.206 0.035 0.127 0.170 0.617 a Unconstrained estimation means that impact of NTMs on trade is not restricted in the econometric estimation. b Constrained estimation means that NTMs are constrained to have a non positive impact on trade in the estimation.

 34

Table 2. AVEs of trade-reducing and trade-facilitating NTMs, by HS section

HS section codes

HS section names

Share of trade-reducing

in NTM- ridden observations

Mean AVE trade-reducing

NTMs (AVE>0)

Mean AVE trade-facilitating NTMs (AVE0)

I Live animals, animal products 0.606 1.204 -0.826 II Vegetable products 0.579 1.047 -0.873 III Fats and oils 0.653 1.315 -0.824 IV Prepared foodstuffs, beverages,

spirits, tobacco 0.570 1.130 -0.803 V Minerals 0.523 1.279 -0.847 VI Chemicals, allied industries 0.352 1.177 -0.871 VII Plastics, rubber 0.552 1.067 -0.818 VIII Hides, leather, furskins 0.530 1.081 -0.899 IX Wood and wood articles 0.598 0.899 -0.828 X Pulp of wood, paper, printing 0.504 0.950 -0.825 XI Textiles, apparel 0.488 1.113 -0.834 XII Footwear, headgear 0.597 1.214 -0.807 XIII Stone, cement, ceramic articles,

glass 0.567 1.141 -0.831 XIV Pearls, precious metals and stones 0.364 0.703 -0.974 XV Base metals and articles 0.532 0.971 -0.828 XVI Machinery, electrical and video

equipment 0.606 1.088 -0.810 XVII Vehicles, aircraft, vessels 0.431 1.248 -0.815 XVIII Optical, photo., medical instr. 0.503 1.052 -0.858 XIX Arms, ammunition 0.183 0.748 -0.931 XX Miscellaneous (furniture, toys,

others) 0.592 1.281 -0.881 All sections 0.518 1.111 -0.840

   

35 

Table 3. Trade restrictiveness indices, summary statistics

Indices MTRI (Tmerc)

Tmerc Tmccr Tmerc Tmerc TRI (T)

T T T T TRI change(dT)

dT dT dT

Protection tariffs overall protection tariffs overall protection overall protection Estimation constrainedb unconstraineda constrainedb unconstraineda constrainedb unconstraineda Estimates all all signif. all signif. all all signif. all signif. all signif. all signif.

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) All 93 countries

Minimum 0.000 0.014 0.009 -0.454 -0.016 0.000 0.086 0.066 0.025 0.029 0.007 0.004 -0.452 -0.030 Maximum 0.257 0.519 0.393 0.195 0.252 0.572 0.847 0.658 0.938 0.733 0.717 0.432 0.879 0.538

Mean 0.081 0.167 0.133 0.011 0.072 0.141 0.357 0.290 0.255 0.171 0.154 0.102 0.043 0.039 Median 0.072 0.133 0.121 0.019 0.066 0.122 0.350 0.257 0.220 0.129 0.123 0.066 0.027 0.015 Std. dev 0.056 0.110 0.084 0.111 0.053 0.097 0.163 0.135 0.173 0.126 0.139 0.095 0.154 0.079

OECD countries Minimum 0.008 0.023 0.021 -0.150 0.008 0.042 0.094 0.071 0.025 0.049 0.009 0.005 -0.123 -0.008 Maximum 0.151 0.303 0.267 0.162 0.146 0.505 0.589 0.549 0.938 0.505 0.347 0.301 0.879 0.255

Mean 0.040 0.102 0.081 -0.003 0.034 0.110 0.342 0.251 0.328 0.130 0.133 0.073 0.080 0.021 Median 0.027 0.083 0.062 0.003 0.021 0.069 0.346 0.239 0.338 0.092 0.120 0.057 0.055 0.005 Std. dev 0.035 0.064 0.055 0.059 0.032 0.101 0.128 0.101 0.195 0.104 0.090 0.065 0.180 0.049

LDCs Minimum 0.030 0.042 0.032 -0.194 0.006 0.049 0.086 0.066 0.072 0.049 0.007 0.004 -0.123 -0.030 Maximum 0.177 0.468 0.351 0.124 0.154 0.225 0.678 0.595 0.255 0.316 0.460 0.354 0.065 0.100

Mean 0.104 0.184 0.149 0.047 0.089 0.132 0.287 0.241 0.171 0.140 0.113 0.084 0.012 0.020 Median 0.097 0.148 0.118 0.076 0.095 0.113 0.237 0.204 0.186 0.121 0.056 0.042 0.022 0.014 Std. dev 0.044 0.130 0.094 0.098 0.043 0.054 0.183 0.170 0.068 0.074 0.143 0.112 0.051 0.031

BRICs Minimum 0.102 0.205 0.173 0.013 0.092 0.125 0.365 0.325 0.216 0.150 0.133 0.106 -0.008 0.022 Maximum 0.257 0.317 0.305 0.172 0.252 0.297 0.668 0.658 0.601 0.572 0.446 0.432 0.361 0.328

Mean 0.150 0.266 0.223 0.081 0.140 0.188 0.485 0.440 0.359 0.264 0.248 0.210 0.117 0.102 Median 0.120 0.270 0.206 0.069 0.109 0.166 0.453 0.388 0.261 0.168 0.207 0.152 0.058 0.028 Std. dev 0.073 0.050 0.058 0.067 0.076 0.081 0.132 0.151 0.210 0.206 0.139 0.151 0.166 0.151 a Unconstrained estimation means that impact of NTMs on trade is not restricted in the econometric estimation. b Constrained estimation means that NTMs are constrained to have a non positive impact on trade in the estimation. OECD: all OECD members included in our sample. BRICs: Brazil, Russia, India and China. LDCs: Burkina Faso, Bangladesh, Ethiopia, Madagascar, Mali, Malawi, Rwanda, Sudan, Senegal, Uganda, Zambia.

   

36 

Figure 1. The impact of NTMs on demand, supply and imports

m(wp) 

wp+τ 

wp+τ+t(NTM) 

x        x’ 

             y 

x, y, m 

p 

wp 

m(wp+τ+total AVE(NTM)) m(wp+τ+t(NTM)) 

 37

Figure 2. Scattered plot of HS6 level NTM AVES averaged over countries and shown by HS2 line

 38

‐1.2

‐1.0

‐0.8

‐0.6

‐0.4

‐0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

2.2

2.4

2.6

2.8

3.0

3.2

3.4

3.6

3.8

4.0

4.2

0 2 4 6 8 10

12

14

16

18

20

22

24

26

28

30

32

34

36

38

40

42

44

46

48

50

52

54

56

58

60

62

64

66

68

70

72

74

76

78

80

82

84

86

88

90

92

94

96

NTM

 AVE  average

 per HS6

 line

 show

n by

 HS2

 line

HS2 line

average of the HS6 averages by HS2 line median of the HS6 averages by HS2  line

Live

 animals, animal 

produ

cts

Vegtable prod

ucts A

nimalandvegetable fats & oils

Prepared foods, bevrages, 

spirits, tobacco produ

c ts

Chem

icals, 

pharmaceu

ticals, 

cosm

etics

Minerals and

 produ

cts

Rubb

er and

 plastics

Hide, leathe

r and

 produ

cts

woo

d and articles, 

Woo

d pu

lp, paper, prin

ted good

s

Textiles, app

arel and

 clothing

Footwear, headgear, umbrellas

articles o

f stone

,cem

ent, ce

ramic  &

 glass

Pearls, precious stones& metals

Basic metals and

 articles of base 

metal

Machinery, m

echanical appliances, electrical equipm

ent

Veh

icles, aircraft, vessels & transport 

Optical, photographic, precision

 & m

edical 

Arm

s and ammunition Furnitures, toys,m

iscellaneous 

 

Figure 3. Mean and median (by HS2) of HS6 NTM AVEs average

  1

Online Appendix (for review only)

Table A.1. Frequency ratios and AVEs of NTMs, by HS 2-digit sector

HS sections

HS2 codes

HS2 names Simple freq.

ratio of NTMs

AVE of NTMs all HS6 lines (mean)

AVE of NTMs if NTM=1 (mean)

Unconstr.

estimation*Constrained estimation

Unconstr. estimation*

Constrained estimation

I 01 Live animals 0.507 0.157 0.349 0.310 0.688 02 Meat & edible meat offal 0.502 0.351 0.496 0.699 0.988 03 Fish and crustaceans 0.451 0.013 0.267 0.028 0.591 04 Dairy products, eggs 0.528 0.501 0.564 0.949 1.069 05 Products of animal origin 0.244 -0.012 0.106 -0.047 0.435

II 06 Live trees & other plans, bulbs, roots 0.489 -0.087 0.125 -0.178 0.255 07 Edible vegetables 0.489 0.119 0.290 0.242 0.592 08 Edible fruit and nuts 0.507 0.177 0.353 0.349 0.698 09 Coffee, tea, maté 0.428 0.088 0.288 0.205 0.673 10 Cereals 0.425 0.055 0.311 0.130 0.731 11 Products of the milling industry 0.371 0.247 0.298 0.665 0.804 12 Oil seeds & oleaginous fruits 0.343 0.038 0.214 0.110 0.625 13 Lac, gums & resins 0.309 -0.164 0.053 -0.530 0.173 14 Vegetable plaiting materials 0.158 0.130 0.149 0.827 0.944

III 15 Animal or vegetable fats and oils 0.370 0.212 0.348 0.573 0.942 IV 16 Preparations of meat, of fish 0.525 0.106 0.302 0.202 0.576

17 Sugars 0.463 0.191 0.316 0.411 0.682 18 Cocoa 0.414 0.083 0.268 0.201 0.647 19 Preparations of cereals, flour, starch or milk 0.452 0.344 0.510 0.762 1.128 20 Preparations of vegetables, fruit, nuts 0.452 0.184 0.354 0.406 0.784 21 Miscellaneous edible preparations 0.500 0.287 0.424 0.574 0.849 22 Beverages, spirits and vinegar 0.361 -0.072 0.180 -0.199 0.499 23 Residues and waste from the food industries 0.200 -0.011 0.125 -0.054 0.625 24 Tobacco 0.466 -0.001 0.223 -0.002 0.478

V 25 Salt 0.084 0.012 0.055 0.147 0.650 26 Ores, slag and ash 0.047 0.014 0.028 0.293 0.584 27 Mineral fuels, mineral oils 0.163 0.062 0.135 0.382 0.831

VI 28 Inorganic chemicals 0.149 -0.005 0.082 -0.035 0.549 29 Organic chemicals 0.195 -0.036 0.089 -0.184 0.455 30 Pharmaceutical products 0.527 -0.101 0.234 -0.191 0.444 31 Fertilizers 0.281 -0.043 0.125 -0.154 0.446 32 Tanning or dyeing extracts 0.167 0.035 0.110 0.209 0.658 33 Essential oils and resinoids 0.287 -0.118 0.085 -0.409 0.296 34 Soaps 0.232 -0.071 0.080 -0.305 0.347 35 Albuminoidal substances 0.203 -0.119 0.038 -0.586 0.188 36 Explosives 0.201 0.023 0.134 0.112 0.667 37 Photographic or cinematographic goods 0.107 0.016 0.068 0.152 0.636 38 Miscellaneous chemical products 0.162 -0.034 0.066 -0.212 0.411

VII 39 Plastics and articles 0.162 0.041 0.104 0.251 0.640 40 Rubber and articles 0.155 0.026 0.106 0.168 0.686

VIII 41 Raw hides and skins 0.117 0.030 0.085 0.253 0.722 42 Leather 0.147 0.018 0.081 0.122 0.553 43 Fur skins and artificial fur 0.102 -0.003 0.033 -0.033 0.319

IX 44 Wood and articles of wood 0.171 0.022 0.087 0.131 0.509 45 Cork and articles 0.107 0.152 0.156 1.422 1.454

  2

46 Straw 0.113 0.006 0.029 0.050 0.257 X 47 Pulp of wood 0.090 0.028 0.059 0.308 0.653 48 Paper 0.137 -0.002 0.064 -0.012 0.465 49 Printed books, newspapers 0.134 0.054 0.067 0.405 0.501

XI 50 Silk 0.175 0.086 0.154 0.490 0.882 51 Wool 0.246 0.093 0.182 0.378 0.739 52 Cotton 0.257 0.020 0.145 0.079 0.562 53 Other vegetable textile fibres 0.219 0.068 0.134 0.308 0.609 54 Man-made filaments 0.303 0.002 0.146 0.005 0.482 55 Man-made staple fibres 0.279 0.042 0.152 0.151 0.546 56 Wadding 0.289 0.055 0.193 0.192 0.668 57 Carpets 0.258 0.113 0.238 0.439 0.922 58 Special woven fabrics 0.242 0.083 0.184 0.345 0.760

59 Impregnated, coated, covered or laminated textile fabrics

0.259 0.091 0.196 0.352 0.759

60 Knitted or crocheted fabrics 0.256 0.079 0.182 0.310 0.710 61 Apparel & clothing accessories, knitted/ crocheted 0.286 0.024 0.153 0.083 0.535 62 Apparel & clothing access., not knitted/ crocheted 0.321 -0.019 0.155 -0.060 0.483 63 Other made-up textile articles 0.273 0.009 0.169 0.033 0.620

XII 64 Footwear 0.362 0.068 0.188 0.187 0.518 65 Headgear 0.130 0.207 0.230 1.587 1.764 66 Umbrellas 0.097 -0.007 0.032 -0.077 0.332 67 Feathers 0.088 0.132 0.146 1.494 1.659

XIII 68 Stone articles 0.087 0.017 0.058 0.198 0.668 69 Ceramic products 0.128 0.032 0.073 0.250 0.568 70 Glass articles 0.119 0.044 0.088 0.365 0.743

XIV 71 Pearls, precious stones and metals 0.015 -0.005 0.004 -0.364 0.273 XV 72 Iron & steel 0.123 0.001 0.064 0.008 0.520

73 Articles of iron or steel 0.147 0.021 0.077 0.141 0.523 74 Copper 0.092 -0.008 0.045 -0.090 0.490 75 Nickel 0.047 0.025 0.042 0.533 0.893 76 Aluminum 0.128 -0.004 0.048 -0.031 0.376 78 Lead 0.050 0.009 0.035 0.184 0.701 79 Zinc 0.086 0.018 0.060 0.210 0.700 80 Tin 0.063 -0.006 0.023 -0.091 0.359 81 Other base metals 0.058 0.066 0.079 1.123 1.354 82 Tools 0.150 0.043 0.090 0.286 0.598 83 Miscellaneous articles of base metal 0.132 0.023 0.078 0.174 0.590

XVI 84 Nuclear reactors 0.167 0.062 0.121 0.373 0.723 85 Electrical machinery & equipment 0.186 0.052 0.120 0.280 0.647

XVII 86 Railway 0.078 0.082 0.094 1.055 1.204 87 Vehicles 0.277 -0.005 0.143 -0.018 0.515 88 Aircraft 0.122 0.007 0.071 0.059 0.586 89 Ships, boats 0.080 0.011 0.044 0.138 0.547

XVIII 90 Optical, photog., measuring, prec., medical instr. 0.184 0.017 0.103 0.091 0.559 91 Clocks and watches 0.000 - - - - 92 Musical instruments 0.068 0.022 0.043 0.317 0.639

XIX 93 Arms and ammunitions 0.306 -0.191 0.057 -0.625 0.186 XX 94 Furniture 0.149 0.127 0.174 0.853 1.171

95 Toys 0.162 0.042 0.106 0.262 0.656 96 Miscellaneous manufactured articles 0.126 0.019 0.068 0.150 0.542

*Unconstrained estimation means that impact of NTMs on trade is not restricted in the econometric estimation. Constrained estimation means that NTMs are constrained to have a non positive impact on trade in the estimation.

  3

Table A.2. AVEs of trade-reducing and trade-facilitating NTMs, by HS 2-digit sector

HS2 codes

HS2 names Share of trade-

reducing in NTM- ridden observations

Mean AVE trade-reducing

NTMs (AVE>0)

Mean AVE trade-facilitating NTMs (AVE0)

01 Live animals 0.628 1.026 -0.901 02 Meat & edible meat offal 0.706 1.301 -0.746 03 Fish and crustaceans 0.436 1.181 -0.862 04 Dairy products, eggs 0.860 1.223 -0.734 05 Products of animal origin 0.431 0.909 -0.770 06 Live trees & other plans, bulbs, roots 0.507 0.441 -0.815 07 Edible vegetables 0.639 0.863 -0.857 08 Edible fruit and nuts 0.559 1.317 -0.876 09 Coffee, tea, maté 0.544 1.095 -0.856 10 Cereals 0.543 0.935 -0.828 11 Products of the milling industry 0.853 0.928 -0.861 12 Oil seeds & oleaginous fruits 0.438 1.422 -0.912 13 Lac, gums & resins 0.258 0.636 -0.935 14 Vegetable plaiting materials 0.800 1.264 -0.920 15 Animal or vegetable fats and oils 0.653 1.315 -0.824 16 Preparations of meat, of fish 0.620 0.857 -0.866 17 Sugars 0.701 0.917 -0.773 18 Cocoa 0.549 1.005 -0.775 19 Preparations of cereals, flour, starch or milk 0.583 1.763 -0.638 20 Preparations of vegetables, fruit, nuts 0.629 1.114 -0.795 21 Miscellaneous edible preparations 0.628 1.341 -0.723 22 Beverages, spirits and vinegar 0.303 1.464 -0.921 23 Residues and waste from the food industries 0.443 0.922 -0.828 24 Tobacco 0.451 0.933 -0.772 25 Salt 0.502 1.194 -0.908 26 Ores, slag and ash 0.682 0.848 -0.900 27 Mineral fuels, mineral oils 0.506 1.513 -0.775 28 Inorganic chemicals 0.379 1.320 -0.862 29 Organic chemicals 0.343 1.198 -0.906 30 Pharmaceutical products 0.319 1.260 -0.870 31 Fertilizers 0.340 1.146 -0.824 32 Tanning or dyeing extracts 0.484 1.144 -0.667 33 Essential oils and resinoids 0.255 0.897 -0.856 34 Soaps 0.302 0.890 -0.822 35 Albuminoidal substances 0.210 0.598 -0.901 36 Explosives 0.443 1.338 -0.862 37 Photographic or cinematographic goods 0.500 1.190 -0.886 38 Miscellaneous chemical products 0.362 0.975 -0.885 39 Plastics and articles 0.574 1.033 -0.803 40 Rubber and articles 0.509 1.145 -0.844 41 Raw hides and skins 0.525 1.313 -0.919 42 Leather 0.521 1.096 -0.938 43 Fur skins and artificial fur 0.560 0.536 -0.757 44 Wood and articles of wood 0.574 0.850 -0.837 45 Cork and articles 0.935 1.587 -0.968 46 Straw 0.614 0.478 -0.631

  4

47 Pulp of wood 0.527 1.370 -0.877 48 Paper 0.462 0.939 -0.829 49 Printed books, newspapers 0.739 0.798 -0.708 50 Silk 0.619 1.369 -0.941 51 Wool 0.568 1.250 -0.766 52 Cotton 0.458 1.133 -0.810 53 Other vegetable textile fibres 0.635 0.943 -0.797 54 Man-made filaments 0.482 0.935 -0.860 55 Man-made staple fibres 0.549 0.941 -0.810 56 Wadding 0.502 1.174 -0.800 57 Carpets 0.507 1.653 -0.811 58 Special woven fabrics 0.499 1.568 -0.873

59 Impregnated, coated, covered or laminated textile fabrics

0.572 1.249 -0.848

60 Knitted or crocheted fabrics 0.521 1.400 -0.877

61 Articles of apparel & clothing accessories, knitted/ crocheted

0.512 0.982 -0.860

62 Art. of apparel & clothing accessories, not knitted/ crocheted

0.399 1.130 -0.851

63 Other made-up textile articles 0.413 1.247 -0.821 64 Footwear 0.561 0.961 -0.800 65 Headgear 0.760 2.371 -0.901 66 Umbrellas 0.508 0.569 -0.745 67 Feathers 0.867 1.876 -0.987 68 Stone articles 0.486 1.283 -0.829 69 Ceramic products 0.565 1.022 -0.752 70 Glass articles 0.621 1.127 -0.882 71 Pearls, precious stones and metals 0.364 0.703 -0.974 72 Iron & steel 0.458 1.032 -0.858 73 Articles of iron or steel 0.556 0.863 -0.761 74 Copper 0.409 1.063 -0.888 75 Nickel 0.641 1.358 -0.937 76 Aluminum 0.489 0.800 -0.825 78 Lead 0.512 1.197 -0.877 79 Zinc 0.513 1.207 -0.840 80 Tin 0.545 0.601 -0.922 81 Other base metals 0.775 1.707 -0.890 82 Tools 0.629 0.925 -0.801 83 Miscellaneous articles of base metal 0.601 0.852 -0.850 84 Nuclear reactors 0.621 1.092 -0.807 85 Electrical machinery & equipment 0.578 1.079 -0.816 86 Railway 0.826 1.448 -0.810 87 Vehicles 0.387 1.237 -0.810 88 Aircraft 0.473 1.135 -0.906 89 Ships, boats 0.504 1.055 -0.794

90 Optical, photographic, measuring, precision, medical instr.

0.495 1.057 -0.855

91 Clocks and watches - - - 92 Musical instruments 0.652 0.988 -0.940 93 Arms and ammunitions 0.183 0.748 -0.931 94 Furniture 0.676 1.694 -0.904 95 Toys 0.583 1.086 -0.889 96 Miscellaneous manufactured articles 0.528 1.053 -0.859

  5

Table A.3. Trade restrictiveness indices, by country

Country Tmerc

Tmerc  Tmerc T T T dT dT dT

Tariffs Overall protection Tariffs Overall protection Tariffs Overall protection

Constrained estimation

Unconstr. estimation*,1

Constrained estimation

Unconstr. estimation*

Constrained estimation

Unconstr. estimation*

Albania 0.117 0.123 0.110 0.134 0.150 0.109 0.018 0.022 0.012 Argentina 0.129 0.178 0.080 0.141 0.341 0.221 0.020 0.116 0.049 Australia 0.057 0.126 -0.079 0.095 0.266 0.009 0.071 -0.089 Austria 0.016 0.075 0.019 0.053 0.397 0.394 0.003 0.157 0.155 Belgium 0.021 0.098 0.018 0.067 0.434 0.369 0.005 0.189 0.136

Burkina Faso 0.106 0.152 0.090 0.122 0.257 0.149 0.015 0.066 0.022 Bangladesh 0.177 0.246 0.107 0.225 0.386 0.255 0.050 0.149 0.065

Belarus 0.085 0.167 0.074 0.106 0.312 0.176 0.011 0.097 0.031 Bolivia 0.080 0.144 0.065 0.086 0.268 0.105 0.007 0.072 0.011 Brazil 0.105 0.247 0.080 0.128 0.416 0.216 0.016 0.173 0.047 Brunei 0.141 0.205 0.156 0.572 0.847 0.581 0.327 0.717 0.338

Canada 0.028 0.057 -0.058 0.076 0.174 0.006 0.030 -0.064 Switzerland 0.040 0.066 -0.072 0.192 0.272 0.037 0.074 -0.055

Chile 0.069 0.107 0.011 0.069 0.195 0.005 0.038 -0.036 China 0.135 0.205 0.013 0.203 0.365 0.041 0.133 -0.008

Ivory Coast 0.094 0.318 -0.340 0.118 0.524 0.014 0.275 -0.256 Cameroon 0.140 0.165 0.137 0.160 0.226 0.186 0.026 0.051 0.034 Colombia 0.112 0.239 -0.003 0.131 0.443 0.693 0.017 0.197 0.481 Costa Rica 0.040 0.042 0.010 0.072 0.096 0.005 0.009 -0.019

Czech Rep. 0.041 0.048 0.002 0.063 0.094 0.004 0.009 -0.023 Germany 0.014 0.068 -0.003 0.049 0.358 0.279 0.002 0.128 0.078 Denmark 0.017 0.110 -0.050 0.047 0.493 0.379 0.002 0.243 0.143

Algeria 0.131 0.392 -0.071 0.160 0.582 0.026 0.339 -0.007 Egypt 0.128 0.421 -0.121 0.197 0.691 0.264 0.039 0.477 0.070 Spain 0.015 0.078 -0.020 0.055 0.494 0.394 0.003 0.244 0.156

Estonia 0.009 0.023 0.003 0.050 0.127 0.002 0.016 -0.001 Ethiopia 0.136 0.148 0.075 0.182 0.217 0.033 0.047 -0.003 Finland 0.011 0.042 -0.003 0.042 0.252 0.166 0.002 0.064 0.028 France 0.013 0.077 -0.002 0.044 0.347 0.243 0.002 0.120 0.059 Gabon 0.153 0.153 0.123 0.175 0.176 0.074 0.031 0.031 0.005

Great Britain 0.019 0.081 -0.005 0.090 0.379 0.275 0.008 0.143 0.076 Ghana 0.144 0.185 0.120 0.245 0.354 0.245 0.060 0.126 0.060

Greece 0.012 0.065 0.026 0.049 0.546 0.507 0.002 0.298 0.258 Guatemala 0.068 0.171 -0.036 0.096 0.357 0.009 0.128 -0.037 Hong Kong 0.000 0.014 -0.042 0.000 0.108 0.000 0.012 -0.038 Honduras 0.067 0.083 0.075 0.092 0.152 0.138 0.008 0.023 0.019 Hungary 0.061 0.113 0.036 0.087 0.249 0.082 0.008 0.062 0.007 Indonesia 0.046 0.082 0.050 0.085 0.355 0.150 0.007 0.126 0.023

India 0.257 0.317 0.172 0.297 0.668 0.601 0.088 0.446 0.361 Ireland 0.008 0.040 0.013 0.042 0.234 0.180 0.002 0.055 0.032 Iceland 0.029 0.061 0.012 0.122 0.231 0.094 0.015 0.053 0.009

Italy 0.017 0.088 0.008 0.072 0.433 0.347 0.005 0.187 0.121 Jordan 0.120 0.262 -0.033 0.163 0.421 0.027 0.177 -0.046 Japan 0.078 0.299 0.162 0.323 0.589 0.473 0.105 0.347 0.224

Kazakhstan 0.043 0.149 0.016 0.073 0.350 0.057 0.005 0.123 0.003

  6

Kenya 0.119 0.127 0.110 0.184 0.206 0.178 0.034 0.043 0.032 South Korea 0.107 0.108 0.107 0.505 0.511 0.510 0.255 0.261 0.261

Lebanon 0.057 0.196 0.042 0.098 0.387 0.175 0.010 0.150 0.031 Sri Lanka 0.074 0.075 0.065 0.138 0.139 0.100 0.019 0.019 0.010 Lithuania 0.021 0.056 -0.052 0.064 0.187 0.004 0.035 -0.058

Latvia 0.027 0.133 0.006 0.073 0.330 0.073 0.005 0.109 0.005 Morocco 0.228 0.471 -0.109 0.275 0.727 0.339 0.075 0.528 0.115 Moldova 0.047 0.072 0.041 0.202 0.239 0.182 0.041 0.057 0.033

Madagascar 0.030 0.042 0.023 0.049 0.107 0.002 0.011 -0.005 Mexico 0.151 0.303 0.025 0.211 0.490 0.235 0.045 0.240 0.055

Mali 0.097 0.129 0.076 0.112 0.183 0.072 0.012 0.034 0.005 Mauritius 0.122 0.207 0.106 0.233 0.386 0.254 0.054 0.149 0.065 Malawi 0.098 0.149 0.112 0.130 0.243 0.165 0.017 0.059 0.027 Malaysia 0.063 0.315 -0.327 0.269 0.560 0.073 0.314 -0.268 Nigeria 0.221 0.419 -0.180 0.309 0.620 0.096 0.384 -0.026

Nicaragua 0.049 0.133 -0.028 0.079 0.294 0.006 0.086 -0.037 Netherlands 0.014 0.083 0.010 0.059 0.483 0.414 0.003 0.233 0.171

Norway 0.045 0.078 0.019 0.255 0.333 0.245 0.065 0.111 0.060 New Zealand 0.027 0.141 -0.150 0.044 0.397 0.002 0.157 -0.091

Oman 0.117 0.176 0.118 0.257 0.375 0.282 0.066 0.140 0.079 Peru 0.126 0.225 0.073 0.129 0.392 0.218 0.017 0.153 0.047

Philippines 0.037 0.276 -0.360 0.068 0.502 0.005 0.252 -0.313 Pap. N. Guinea 0.029 0.093 0.008 0.152 0.292 0.078 0.023 0.085 0.006

Poland 0.103 0.144 0.031 0.150 0.270 0.023 0.073 -0.002 Portugal 0.036 0.131 0.039 0.175 0.452 0.338 0.031 0.205 0.114 Paraguay 0.107 0.200 0.015 0.123 0.386 0.055 0.015 0.149 0.003 Romania 0.120 0.176 0.116 0.157 0.300 0.216 0.025 0.090 0.047 Russia 0.102 0.293 0.057 0.125 0.489 0.261 0.016 0.240 0.068

Rwanda 0.088 0.130 0.124 0.113 0.237 0.219 0.013 0.056 0.048 Saudi Arabia 0.142 0.158 0.062 0.348 0.368 0.248 0.121 0.135 0.062

Sudan 0.174 0.468 -0.077 0.214 0.678 0.222 0.046 0.460 0.049 Senegal 0.086 0.374 -0.194 0.108 0.556 0.012 0.309 -0.123 Singapore 0.000 0.222 -0.454 0.000 0.428 0.000 0.183 -0.452

El Salvador 0.064 0.133 0.026 0.096 0.269 0.009 0.072 -0.021 Slovenia 0.102 0.197 -0.048 0.120 0.346 0.015 0.120 -0.049 Sweden 0.014 0.059 -0.017 0.052 0.254 0.025 0.003 0.064 0.001 Thailand 0.109 0.132 0.083 0.168 0.248 0.144 0.028 0.062 0.021

Trinidad & T. 0.072 0.082 0.069 0.296 0.315 0.301 0.088 0.099 0.091 Tunisia 0.228 0.363 0.099 0.300 0.525 0.357 0.090 0.276 0.128 Turkey 0.043 0.105 -0.001 0.095 0.259 0.938 0.009 0.067 0.879 Tanzania 0.137 0.519 0.083 0.160 0.809 0.572 0.026 0.655 0.327 Uganda 0.067 0.067 0.065 0.084 0.086 0.079 0.007 0.007 0.006 Ukraine 0.064 0.285 0.195 0.159 0.519 0.437 0.025 0.270 0.191 Uruguay 0.097 0.208 0.028 0.117 0.408 0.203 0.014 0.166 0.041

United States 0.024 0.083 -0.138 0.049 0.256 0.002 0.065 -0.123 Venezuela 0.135 0.231 0.017 0.158 0.383 0.034 0.025 0.147 0.001

South Africa 0.069 0.077 0.050 0.131 0.157 0.044 0.017 0.025 0.002 Zambia 0.086 0.116 0.115 0.113 0.205 0.207 0.013 0.042 0.043

*Unconstrained estimation means that impact of NTMs on trade is not restricted in the econometric estimation. Constrained estimation means that NTMs are constrained to have a non positive impact on trade in the estimation. 1 With an externality and some negative AVEs, the MTRI can be smaller or larger than the TRI and the two indices may not have similar signs. : LDCs. : BRIC countries. : OECD countries.